Been seeing a lot of confusion on this lately, especially for folks just getting started. Both checkpoints and LoRAs change your outputs, but they work in totally different ways under the hood.
Think of a checkpoint as a whole new artist. It's a complete model with its own style, knowledge, and "brain." Switching checkpoints is like hiring a watercolorist instead of a photorealistic painter. A LoRA, on the other hand, is like giving your current artist a specific training course or a new tool. It's a small add-on that teaches the model a new concept (like a specific character face) or style without replacing everything else. LoRAs are lightweight and stackable, so you can mix a few to get a unique combo. For a real-world analogy: swapping checkpoints is like getting a new computer; using a LoRA is like installing a new, specialized piece of software on your existing one.
Hope that clears it up! What's been your experience mixing them?
Trust the trial period.